Research map / Architectures and efficiency / CNN design
ChannelNets: Compact and Efficient Convolutional Neural Networks via Channel-Wise Convolutions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use group convolutions with group channel-wise convolutions for information fusion (ablation) | ChannelNet-v1 · ImageNet ILSVRC 2012 validation | Top-1 accuracy: 0.697 -> 0.705 | Helped |
| Replace depth-wise separable convolutions with group modules and group channel-wise modules in deeper layers | ChannelNet-v1 · ImageNet ILSVRC 2012 validation | FLOPs: 28.5% reduction in FLOPs | Helped |
| Replace the last depth-wise separable convolution with a depth-wise separable channel-wise convolution | ChannelNet-v2 · ImageNet ILSVRC 2012 validation | Top-1 accuracy: 0.684 -> 0.695 | Helped |
| Replace the fully connected classifier with a convolutional classification layer | ChannelNet-v3 · ImageNet ILSVRC 2012 validation | Top-1 accuracy: 0.637 -> 0.667 | Helped |
| Apply width multiplier 0.75 across ChannelNet-v1 layers | ChannelNet-v1 · ImageNet ILSVRC 2012 validation | Top-1 accuracy: 0.705 -> 0.678 | Hurt |
| Apply width multiplier 0.5 across ChannelNet-v1 layers | ChannelNet-v1 · ImageNet ILSVRC 2012 validation | Top-1 accuracy: 0.705 -> 0.627 | Hurt |
Closest papers by meaning
- Training CNNs with Selective Allocation of Channels (ICML 2019 · similarity 0.88)
- Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers (ICLR 2018 · similarity 0.88)
- Approximation and non-parametric estimation of ResNet-type convolutional neural networks (ICML 2019 · similarity 0.87)
- Rethinking Channel Dimensions for Efficient Model Design (CVPR 2021 · similarity 0.87)
- CT-Net: Channel Tensorization Network for Video Classification (ICLR 2021 · similarity 0.87)
- MobileODE: An Extra Lightweight Network (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.